Scribd
Upload
24 views

Machine Learning

Machine learning unit 2

Uploaded by

hemanthonlyking
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
24 views

Machine Learning

Machine learning unit 2

Uploaded by

hemanthonlyking
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 9
UNIT bas ” Medeting and Evaluation & Basics ob Teature Ergtincering = Thtedutn: ing. asic learning procese of a Machine (& divided into “Thee ne > DF Data “Trput ~ 2) Absvauion r 3) * Genuatization PP wire ccteane clepavtment ol new city polfce has got. a0 infor mation thet Q Giminal % going bo taunch 47 attous ON the main condidate for YPeming eteccion ental => sleiting a moda: Th Machine leaning prradigen , the pot causes of Afchubance - " | Gample +s Averace inom ot the lecal population, wxa pon SACS, «, PP os MEO OL imigrank are input variables. are_aléo caued _ “S Preciictorsy attsbutes ;features , independent: voricles (OF) - Simply Nasiables , : ; ees The Number of- criminal inddens. & an sutpuk Variable . ALO Gilled 03 Response 0% dependent Veviable , Toput variables = aN be denoted by MOK el. Xn) end output nyautable ts Sli fy sym yet 2) = * “The telatenghip techofen X andl y ie y=to0re 1. Uobete, P ie betrget- function and P- @ ig vardory emor value re Thre cite Three Yypes of functions - [? est funtion ~ ( we meciide hou, fhany Warne Fete Y7 = vVvVy * Lose funuion Conly one - thet is cohin fs Precrced wwi0 % Objective Pinan — Pcoherhes TE Ker bon * cost function ic also Called as em function”, : Thae are three byocd categories of Mathine learning eppreathes : @ Lupavised => * classi teatton * Regresston © unsupavised =) “* Clustevin ¥% Assodctron Prralysrs . © Reistorcement: eel Tee gy crower Bet = = _ ~~ pony] 22upuyy wosnpung ZULLIVUvUe " predtdve Modes + corte classikeahon ord Requ fodet Representation ¢ “Toker pretabil rf tep ie + { \ High variance ‘ ig: Bias- variance bade— Eiawati fpafrmance_ oka model = ~ Superiised leaning —claseifzaton : ; sk Th supeyised leami 7 Oe major task is classiWcation , Th respons! ili ot ee et assign class label to the benget feake bastd on the Value of predicred Jeahsres , For examples T the | problem of predicting exin/ Joss in a erke marth the classthia upift ausign a value votn/ioes bo the target feahw “Thee cle by passiotlittes With “reed to Cute math | ; @) The model predicted ein; and the team soon C Tp). CTrue positive) b) We meded predfetecl voiny ond: the team’ lost Grp) C False positive) Q “The medel preatited teste and team won ( FN) Cfalee Negaive) 4) “the mode predfled fost and team lat cm) Ctme Negative), ~The First: case, tHe, the model preatiked we win 2nd the team vor) fs @ case wwhese, the model hae ConceHy clasihed data instances at the clace of interes These cates are refered as Tine pesthive Tp) caes. — —“The kecond cae, Ce, the “model pede] win ond team lott & a case Ushee the model fnconectiy clacsifed cater inchancet 8 Ye clase of trleretl. these cares are” vefened ae Falte Positive (pp ) cases. pied fled oultome Adue! oubeeme Rqt retails of Mode claxefReakion. Model Accuracy + Model Accuracy ie eqiven by bok no Lb tome | cloceification CDITN) Atvidedl by toted no-at clasdficaki one done. GConfusfon mahix + A mehix lontaining comet and tnconet= prectitene in the hrm ob Tp, Fer TH, PA ff Known oF Corkuston matele . “The win ovloss prediction of ufucet match has bao classes ce Ynterete. fr that reaton pte will generate 2x2 confusion mabnX . Pretuel wit) | Aatual loss Preatucel win es C71) 4 CR) predived os [a crs) “| 4 cn) | bi Medel accu! = t =) Meaney” RAN = B44 Say < TRH FINTEN Go ugaeg Jo ty Grorrake= Fpre ‘pr pea EE = b_<6/: @yuqitq — 100 Gro rake = |— model acces : = 1 ay = 0-06 ( sf) Kappa Value 2 “tk-indteates the adjusted model acavay “TEfs Caleataked by 4Si9g the formula. Kappa value CK) = pee) = pr) “l= porn twvVvuyPvUvu~,weYY ~ Sundaram Finance Limited ‘ where pla)= pryertion ck obsewed agreement: bo attual and prectictecd i overall Daleset « Pla) = TPH Se, = W-o-y, HEB TINGEN gery Tade 1° PUps)= proportion ok expeured, aqnecmen = blo actual and prectized d. PAP TPA PAIN eet “IPATPETNTIN pePpAINHAY app HAVIN’ ” Ap tei mip pl Here; Tp=s5, TN=9 Fp=u,; FN=2 =. 85+ X $644 1 at4 = et . S5tutse B5+Ntt ese yy ay bout ate (pri= 89 ge wu YB mp 100 * Tao” Foo Xoo = 0-84 KO-kF Hor KOS = 0-4886 = o-qy—o-tess © (—o0-AkBe = OFmre 4 "i Kappa value can be \ flat the mex johich, ‘vepretente perfect: age Ho models predation ond attual valde . Seosibiity: “Te measuics the proportion of Tp examples (ov) posible cates conith wee somettly claesikies Sensitivity = TPS 85 2 861 wig ata, et Ph, TPTFN Beye BF yoo spritaty - ° [S444 De measuies the peeree de negative exGmplet uohich dle. Conertty elauikes . opeiftaty = “IN = 4H Le notre 6124, i TN EP ty 8 spressient ap eR 185 gags o AS-S f Tethp Boty 84 J = 85 = 0-494 = qa: — Recau \ ; pen ot “Te tawee the: brarment SEP meacure ¢ Te Combines the presen ard Recall mean of predsicn Ard Recall fon & vecall Fo meawie = 2% pw prevsion cp recall F Dx OGSMONTT 2 17866, = o:1891 a OF 1 LOTT ey 14 Roc cuwes [ Recciver eperaé charencteristic Bie Te hepe in visualizing the performance of a model - SP pe gern = Ck TPTPN hee pn = The foucwing fequree shows the cule shows boo clarcibers ? ciestify 1 and elosily 2. {- a Model t > mere) CB mance Ol + i { SEY \ wy Erp Vaue at : Bidet aianci eae 7 ob: “assther a Vv ‘ 0-44 sop, the The? E] on “Te Clossibed ot wilh no prcalie Vane” under Gurle (nud) t ae 516. or8 Bg Sib. ok Re ch) rpc dfaly ) Ponty Pouvuyy wp.opung ete

You might also like